File size: 11,210 Bytes
297c44b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 |
import sys
import torch
import speechbrain as sb
from speechbrain.utils.distributed import run_on_main
from hyperpyyaml import load_hyperpyyaml
class ASR(sb.core.Brain):
def compute_forward(self, batch, stage):
batch = batch.to(self.device)
wavs, wavs_len = batch.sig
tokens_bos, _ = batch.tokens_bos
feats = self.hparams.compute_features(wavs)
current_epoch = self.hparams.epoch_counter.current
feats = self.modules.normalize(feats, wavs_len, epoch=current_epoch)
src = self.modules.CNN(feats)
enc_out, pred = self.modules.Transformer(
src, tokens_bos, wavs_len, pad_idx=self.hparams.pad_index
)
logits = self.modules.ctc_lin(enc_out)
p_ctc = self.hparams.log_softmax(logits)
pred = self.modules.seq_lin(pred)
p_seq = self.hparams.log_softmax(pred)
hyps = None
if stage == sb.Stage.TRAIN:
hyps = None
elif stage == sb.Stage.VALID:
hyps = None
current_epoch = self.hparams.epoch_counter.current
if current_epoch % self.hparams.valid_search_interval == 0:
hyps, _ = self.hparams.valid_search(enc_out.detach(), wavs_len)
elif stage == sb.Stage.TEST:
hyps, _ = self.hparams.test_search(enc_out.detach(), wavs_len)
return p_ctc, p_seq, wavs_len, hyps
def compute_objectives(self, predictions, batch, stage):
(p_ctc, p_seq, wavs_len, hyps,) = predictions
ids = batch.id
tokens_eos, tokens_eos_len = batch.tokens_eos
tokens, tokens_len = batch.tokens
attention_loss = self.hparams.seq_cost(
p_seq, tokens_eos, length=tokens_eos_len
)
ctc_loss = self.hparams.ctc_cost(p_ctc, tokens, wavs_len, tokens_len)
loss = (
self.hparams.ctc_weight * ctc_loss
+ (1 - self.hparams.ctc_weight) * attention_loss
)
if stage != sb.Stage.TRAIN:
current_epoch = self.hparams.epoch_counter.current
valid_search_interval = self.hparams.valid_search_interval
if current_epoch % valid_search_interval == 0 or (
stage == sb.Stage.TEST
):
predictions = [
hparams["tokenizer"].decode_ids(utt_seq).split(" ")
for utt_seq in hyps
]
targets = [
transcription.split(" ")
for transcription in batch.transcription
]
if self.hparams.remove_spaces:
predictions = [
"".join(prediction_words)
for prediction_words in predictions
]
targets = [
"".join(target_words) for target_words in targets
]
self.cer_metric.append(ids, predictions, targets)
self.acc_metric.append(p_seq, tokens_eos, tokens_eos_len)
return loss
def fit_batch(self, batch):
self.check_and_reset_optimizer()
predictions = self.compute_forward(batch, sb.Stage.TRAIN)
loss = self.compute_objectives(predictions, batch, sb.Stage.TRAIN)
(loss / self.hparams.gradient_accumulation).backward()
if self.step % self.hparams.gradient_accumulation == 0:
self.check_gradients(loss)
self.optimizer.step()
self.optimizer.zero_grad()
self.hparams.noam_annealing(self.optimizer)
return loss.detach()
def evaluate_batch(self, batch, stage):
with torch.no_grad():
predictions = self.compute_forward(batch, stage=stage)
loss = self.compute_objectives(predictions, batch, stage=stage)
# origin function is call loss.detach().cpu()
return loss.detach()
def on_stage_start(self, stage, epoch):
if stage != sb.Stage.TRAIN:
self.acc_metric = self.hparams.acc_computer()
self.cer_metric = self.hparams.cer_computer()
def on_stage_end(self, stage, stage_loss, epoch):
stage_stats = {"loss": stage_loss}
if stage == sb.Stage.TRAIN:
self.train_stats = stage_stats
else:
stage_stats["ACC"] = self.acc_metric.summarize()
current_epoch = self.hparams.epoch_counter.current
valid_search_interval = self.hparams.valid_search_interval
if (
current_epoch % valid_search_interval == 0
or stage == sb.Stage.TEST
):
stage_stats["CER"] = self.cer_metric.summarize("error_rate")
if stage == sb.Stage.VALID and sb.utils.distributed.if_main_process():
current_epoch = self.hparams.epoch_counter.current
if current_epoch <= self.hparams.stage_one_epochs:
lr = self.hparams.noam_annealing.current_lr
steps = self.hparams.noam_annealing.n_steps
optimizer = self.optimizer.__class__.__name__
else:
lr = self.hparams.lr_sgd
steps = -1
optimizer = self.optimizer.__class__.__name__
epoch_stats = {
"epoch": epoch,
"lr": lr,
"steps": steps,
"optimizer": optimizer,
}
self.hparams.train_logger.log_stats(
stats_meta=epoch_stats,
train_stats=self.train_stats,
valid_stats=stage_stats,
)
self.checkpointer.save_and_keep_only(
meta={"ACC": stage_stats["ACC"], "epoch": epoch},
max_keys=["ACC"],
num_to_keep=10,
)
elif stage == sb.Stage.TEST:
self.hparams.train_logger.log_stats(
stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
test_stats=stage_stats,
)
with open(self.hparams.cer_file, "w") as cer_file:
self.cer_metric.write_stats(cer_file)
self.checkpointer.save_and_keep_only(
meta={"ACC": 1.1, "epoch": epoch},
max_keys=["ACC"],
num_to_keep=1,
)
def check_and_reset_optimizer(self):
current_epoch = self.hparams.epoch_counter.current
if not hasattr(self, "switched"):
self.switched = False
if isinstance(self.optimizer, torch.optim.SGD):
self.switched = True
if self.switched is True:
return
if current_epoch > self.hparams.stage_one_epochs:
self.optimizer = self.hparams.SGD(self.modules.parameters())
if self.checkpointer is not None:
self.checkpointer.add_recoverable("optimizer", self.optimizer)
self.switched = True
def on_fit_start(self):
"""Initialize the right optimizer on the training start"""
super().on_fit_start()
current_epoch = self.hparams.epoch_counter.current
current_optimizer = self.optimizer
if current_epoch > self.hparams.stage_one_epochs:
del self.optimizer
self.optimizer = self.hparams.SGD(self.modules.parameters())
if self.checkpointer is not None:
group = current_optimizer.param_groups[0]
if "momentum" not in group:
return
self.checkpointer.recover_if_possible(
device=torch.device(self.device)
)
def on_evaluate_start(self, max_key=None, min_key=None):
super().on_evaluate_start()
checkpointers = self.checkpointer.find_checkpoints(
max_key=max_key, min_key=min_key
)
checkpointer = sb.utils.checkpoints.average_checkpoints(
checkpointers, recoverable_name="model", device=self.device
)
self.hparams.model.load_state_dict(checkpointer, strict=True)
self.hparams.model.eval()
def dataio_prepare(hparams):
@sb.utils.data_pipeline.takes("transcription")
@sb.utils.data_pipeline.provides(
"transcription", "tokens_bos", "tokens_eos", "tokens"
)
def transcription_pipline(transcription):
yield transcription
tokens_list = hparams["tokenizer"].encode_as_ids(transcription)
tokens_bos = torch.LongTensor([hparams["bos_index"]] + (tokens_list))
yield tokens_bos
tokens_eos = torch.LongTensor(tokens_list + [hparams["eos_index"]])
yield tokens_eos
tokens = torch.LongTensor(tokens_list)
yield tokens
@sb.utils.data_pipeline.takes("wav")
@sb.utils.data_pipeline.provides("sig")
def audio_pipline(wav):
sig = sb.dataio.dataio.read_audio(wav)
return sig
@sb.utils.data_pipeline.takes("wav")
@sb.utils.data_pipeline.provides("sig")
def sp_audio_pipline(wav):
sig = sb.dataio.dataio.read_audio(wav)
sig = sig.unsqueeze(0)
sig = hparams["speed_perturb"](sig)
sig = sig.squeeze(0)
return sig
datasets = {}
data_folder = hparams["data_folder"]
output_keys = [
"transcription",
"tokens_bos",
"tokens_eos",
"tokens",
"sig",
"id",
]
default_dynamic_items = [transcription_pipline, audio_pipline]
train_dynamic_item = [transcription_pipline, sp_audio_pipline]
for dataset_name in ["train", "dev", "test"]:
if dataset_name == "train":
dynamic_items = train_dynamic_item
else:
dynamic_items = default_dynamic_items
json_path = f"{data_folder}/{dataset_name}.json"
datasets[dataset_name] = sb.dataio.dataset.DynamicItemDataset.from_json(
json_path=json_path,
replacements={"data_root": data_folder},
dynamic_items=dynamic_items,
output_keys=output_keys,
)
return datasets
if __name__ == "__main__":
hparams_file_path, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
sb.utils.distributed.ddp_init_group(run_opts)
with open(hparams_file_path) as hparams_file:
hparams = load_hyperpyyaml(hparams_file, overrides)
sb.create_experiment_directory(
experiment_directory=hparams["output_folder"],
hyperparams_to_save=hparams_file_path,
overrides=overrides,
)
run_on_main(hparams["pretrainer"].collect_files)
hparams["pretrainer"].load_collected(device=run_opts["device"])
datasets = dataio_prepare(hparams)
asr_brain = ASR(
modules=hparams["modules"],
opt_class=hparams["Adam"],
hparams=hparams,
run_opts=run_opts,
checkpointer=hparams["checkpointer"],
)
asr_brain.fit(
asr_brain.hparams.epoch_counter,
datasets["train"],
datasets["dev"],
train_loader_kwargs=hparams["train_dataloader_opts"],
valid_loader_kwargs=hparams["valid_dataloader_opts"],
)
# asr_brain.evaluate(
# datasets["test"],max_key="ACC", test_loader_kwargs=hparams["test_dataloader_opts"]
# )
|